Infrared Laser Chemistry of Trichlorosilane in View of Silicon Isotope

Institute of Spectroscopy RAS, 142190 Troitsk, Moscow region, Russia. J. Phys. Chem. A , 2003, 107 (41), pp 8578–8583. DOI: 10.1021/jp0356362. Publi...
12 downloads 0 Views 93KB Size
8578

J. Phys. Chem. A 2003, 107, 8578-8583

Infrared Laser Chemistry of Trichlorosilane in View of Silicon Isotope Separation M. Polianski, O. V. Boyarkin,* and T. R. Rizzo Laboratoire de chimie physique mole´ culaire, EÄ cole Polytechnique Fe´ de´ rale de Lausanne, CH-1015 Lausanne, Switzerland

V. M. Apatin, V. B. Laptev, and E. A. Ryabov* Institute of Spectroscopy RAS, 142190 Troitsk, Moscow region, Russia ReceiVed: June 10, 2003; In Final Form: July 29, 2003

With a view toward laser isotope separation of Si, we have studied infrared multiphoton dissociation (IRMPD) of room temperature trichlorosilane, SiHCl3. Over the wavelength range investigated, multiphoton dissociation of the room temperature species exhibits a maximum efficiency at 12.6 µm and a threshold fluence of only ∼1 J/cm2. Vibrational overtone preexcitation of SiHCl3 to the first SiH-stretch overtone (2ν1) prior to IRMPD results in a 10-fold increase of the dissociation yield compared to molecules with only thermal excitation. In an effort to collect the nascent SiCl2 dissociation fragments, we have tested a number of different molecules that could serve as a scavenger to convert them into a stable gaseous compound. Several of these molecules react directly with trichlorosilane after being decomposed by collisional energy transfer from vibrationally excited SiHCl3 and therefore are not suitable for a laser isotope separation process. Of the compounds tested, we find that only BCl3 scavenges SiCl2 without significant reaction with the starting material.

I. Introduction The large increase in thermal conductivity of silicon upon isotopic purification1 and the predicted unique properties of silicon isotope superlattices2 suggest significant improvement of semiconductor devices upon use of isotopically enriched Si. Producing isotopically pure semiconductors is, however, currently limited by the high cost of enrichment. The development of economically feasible approaches for isotopic enrichment of these materials would facilitate exploring their potential in practical devices. The method of molecular laser isotope separation (MLIS) based on infrared multiphoton dissociation (IRMPD) has been investigated for the last three decades and applied to many isotopes, and recently it has been implemented for commercial production of carbon-13.3 While isotope separation of silicon by IRMPD of naturally abundant (92.1% 28Si, 4.7% 29Si, 3.2% 30Si) samples of Si(OCH3)4 and ((CH3)3Si)2O,4 CH3SiF3, and C6H5SiF3 5 has achieved a maximum enrichment in 30Si of only 5-15%, this approach has shown some promise for laboratory-scale separation of silicon isotopes using Si2F6 as a working molecule,6-8 with a reported enrichment of 30Si to 46% in the SiF4 dissociation product6 and 20% 29Si in the SiF2 fragment.8 Although this is currently the most advanced laser-based technique for silicon isotope separation, the present level of enrichment is still far below that required for electronic applications. A new highly selective MLIS approach based on overtone preexcitation of a high frequency vibration followed by IRMPD of the preexcited molecules has been recently proposed and studied in view of separating carbon9 and silicon10 isotopes. In both cases, this approach has achieved high isotopic selectivity, allowing enrichment of minor isotopes up to 99% in a singlestage process. For carbon isotope separation, this new technique * Corresponding authors. E-mail: [email protected]; Ryabov@ isan.troitsk.ru.

has been applied to CF3H, which as a symmetric top has a sharp, prominent Q-branch in the C-H stretch overtone bands, allowing efficient preexcitation of the parent molecules. In the case of silicon, this approach has been tested with SiH4, a spherical top with a broad Q-branch in the Si-H overtone bands and a low density of vibrational states. These characteristics, along with the low pressure used in these experiments, resulted in a fairly low estimate for the productivity. To separate silicon isotopes successfully using the overtone-preexcitation infrared multiphoton dissociation (OP-IRMPD) approach, one must find a more suitable parent compound. We report here our investigation of the applicability of the OP-IRMPD technique to the molecule trichlorosilane (SiHCl3). This molecule is a good candidate for such an approach in that it has a light atom stretch vibration for overtone preexcitation (ν1) that exhibits a sharp, ∼1 cm-1, Q-branch.11,12 In the minor isotopic species the Q-branch has been assigned only for the fundamental transitions.13 Extrapolation of these data to the first overtone suggests ∼2.6 cm-1 isotopic shift per mass unit. The 2ν1 Q-branch in the 29SiHCl3 may partially overlap with a “hot” band, while the position of this branch in the 30Si species should allow a significant isotopic selectivity of the preexcitation, especially at low temperatures, when the “hot” bands are substantially supressed. The particular selectivity figures for different preexcitation conditions have to be determined experimentally, perhaps with isotopically enriched SiHCl3. In this work we use IRMPD of naturally abundant SiHCl3 molecules vibrationally preexcited to the first Si-H stretch overtone to determine optimal parameters for selective dissociation of these species relative to the ground-state molecules, since this puts an upper limit for the isotopic selectivity of the process. Performing isotope-selective dissociation of the parent molecule is necessary but not sufficent for a successful isotope separation scheme, however. As in any MLIS process, the dissociation fragment(s) generated by IRMPD carrying the isotope of interest,

10.1021/jp0356362 CCC: $25.00 © 2003 American Chemical Society Published on Web 09/23/2003

Infrared Laser Chemistry of Trichlorosilane

J. Phys. Chem. A, Vol. 107, No. 41, 2003 8579

Figure 1. Schematic energy level diagram for infrared multiphoton dissociation of vibrationally excited SiHCl3 followed by LIF detection of SiCl2.

which in the present case is SiCl2, must be collected and separated from the parent compound. The main reaction pathway for these radicals is formation of solid polymers of the type (SiCl2)n. Collection of this polymeric material would be inconvenient. The use of scavengers to convert radical products of chemical reactions into stable gaseous compounds has already been demonstrated in MLIS experiments.14,15 In the second part of this work, we test the use of several potential scavenger molecules to collect the IRMPD products of SiHCl3. II. Experimental Section Studies of SiHCl3 dissociation by thermal16-18 and laser19 pyrolysis suggest that the lowest energy decomposition channel for trichlorosilane in the electronic ground state is the production of SiCl2 and HCl:

SiHCl3†† f SiCl2 + HCl

(1)

Under certain conditions, the SiCl2 radicals polymerize to form (SiCl2)n.17,18 We measure in situ the relatiVe dissociation yield of reaction 1 after IRMPD using laser-induced fluorescence (LIF) detection of SiCl2 immediately after the dissociation laser pulse. To measure the absolute IRMPD yield of SiHCl3, we monitor its change in concentration by direct absorption after a known number of irradiation pulses. Figure 1 shows a schematic energy level diagram for IRMPD of vibrationally excited SiHCl3 followed by LIF detection of SiCl2. First, an IR laser pulse excites the first overtone band (2ν1) of the Si-H stretch vibration. Approximately 400 ns later, a pulse from an NH3 laser promotes some fraction of the vibrationally excited molecules to energies above the dissociation threshold via infrared multiphoton excitation of the ν4 bending vibration. The SiCl2 dissociation fragments are then detected by a third laser via LIF 600 ns after the preexcitation pulse. By choosing properly the dissociating frequency and fluence, one can dissociate selectively the preexcited molecules without significant dissociation of the ground-state species. The apparatus we employ for these experiments, depicted schematically in Figure 2, has been described in detail elsewhere.20,21 Briefly, we generate 3-4 mJ of infrared radiation at 4450.5 cm-1 for excitation of the first Si-H stretch overtone by difference frequency mixing the output of a Nd:YAG pumped dye laser with the fundamental of the same pump laser in a LiNbO3 crystal. As a dissociating laser we use either a home-

Figure 2. Apparatus for experiments on IRMPD of preexcited SiHCl3.

built NH3 laser pumped by a pulsed CO2 laser or the CO2 laser directly. The dissociating pulse from the NH3 laser largely resembles the pumping pulse, exhibiting a 150-200 ns initial peak and a 3-4 µs long tail. Because we probe the SiCl2 dissociation products by LIF after only the first 600 ns of the dissociating pulse, the parent molecules that we investigate have experienced only 32% of the total fluence. For LIF detection of the SiCl2 fragments at 325.084 nm (via the A ˜ 1B1 (050) r X 1A1 (000) transition22), we use the second harmonic of another Nd:YAG-pumped dye laser. The probe laser beam is combined with the preexcitation beam on a dichroic mirror. The latter is focused by an F ) 60 cm lens, resulting in 1.5 J/cm2 fluence in the beam waist. The two beams are overlapped with the counter-propagating NH3 laser beam, which is focused by an F ) 50 cm lens to the center of the vacuum chamber. Semiconductor grade SiHCl3 (Aldrich) slowly flows through the chamber at constant pressure. In the second part of this work where we concentrate on product collection, experiments on IR-laser-induced chemistry of SiHCl3 in the presence of different scavenger molecules have been performed in small (2.3- and 6.6-cm long) cells equipped with salt windows. The NH3 laser beam is in this case focused by an F ) 10 cm lens, resulting in a fluence of up to 10 J/cm2 at the beam waist. Sharp focusing results in a significant variation of the dissociation fluence, and hence the dissociaion yield, along the length of the cell. For simplicity we use an average dissociation yield, β, to characterize the process. This yield is calculated from the attenuation of the IR absorption bands of SiHCl3 measured in an IR spectrophotometer before and after laser irradiation of the cell. We employ the same technique, IR spectrophotometry, for identification and concentration measurements of the products of chemical reactions between scavenger molecules and the primary dissociation products of SiHCl3. For this we record IR spectra of the gas mixture in the cell before and after irradiation by a certain number of ammonia laser pulses. The difference of the two spectra is compared with reference spectra of the expected chemical products. Most of these reference spectra have been taken from literature, although some of them have been obtained in our laboratory. We define the relative dissociation yield of a particular product as the ratio of its quantity to the quantity of the dissociated trichlorosilane.

8580 J. Phys. Chem. A, Vol. 107, No. 41, 2003

Polianski et al.

Figure 3. IRMPD yield for room-temperature SiHCl3 as a function of the dissociation laser wavenumber (dots, connected by solid line). For comparison, an absorption spectrum of the ν4 band in room-temperature SiHCl3 is shown.

III. Results and Discussion A. IRMPD of Room-Temperature and Vibrationally Preexcited SiHCl3. Our first attempt at IRMPD of SiHCl3 used excitation of the overlapped 2ν2 and ν4 + ν6 bands11 centered at 986 cm-1 by a CO2 laser tuned to the 10R(20) and 10P(20) lines at 982.02 cm-1 and 944.18 cm-1, respectively. Under these conditions, we detected a product LIF signal only when the effective dissociation fluence exceeded 20-30 J/cm2 for a sample pressure in the range of 0.1-2.0 Torr. Such high dissociation threshold fluence is likely due to the low absorption intensity of the pumped band, making this scheme unsuitable for a practical isotope separation process. This led us to use an ammonia laser for pumping the strong ν4 fundamental at 810.8 cm-1. Figure 3 shows IRMPD yields measured for a few lines of the NH3 laser using spectrometric detection of SiHCl3 after the irradiation period. For comparison, an absorption spectrum of the ν4 band is plotted on the same figure. The frequency of maximum yield is shifted by ∼10-15 cm-1 to the lowfrequency side from the maximum of the absorption spectra, prompting us to use the NH3 laser lines below 780 cm-1 for IRMPD of vibrationally preexcited SiHCl3. The overall IRMPD efficiency of SiHCl3 by the ammonia laser is significantly higher than that by the CO2 laser. Despite the fact that the output energy of our ammonia laser on weak red lines (780, 773 cm-1) is only 8-10% of the CO2 laser energy, the dissociation yield is more than an order of magnitude higher than for direct dissociation of trichlorosilane by the latter. This suggests that infrared multiphoton excitation of trichlorosilane is 2 orders of magnitude more efficient when pumping through the ν4 fundamental than that via the 2ν2/(ν4 + ν6) band. We thus concentrate on NH3-laser-induced dissociation of SiHCl3 in studying its suitability for isotope separation. Figure 4 shows the dissociation yield, measured by LIF detection of SiCl2, as a function of the effective dissociation fluence for both vibrationally unexcited SiHCl3 and for SiHCl3 molecules preexcited to the 2ν1 vibrational level. Dissociation of molecules in both vibrational states increases significantly with increasing fluence over the range investigated, but the increase is slower for the preexcited species, in line with our expectations.20 An appreciable number of preexcited molecules dissociates at a fluence below 1 J/cm2, where the dissociation yield from the room-temperature species is still small. The ratio of these two dissociation curves, which we call the dissociation selectivity, is plotted on the same figure (right-hand scale). Along with the isotopic selectivity of the preexcitation step, the dissociation selectivity as defined here limits the isotopic selectivity of the OP-IRMPD approach. We expect infrared multiphoton dissociation of the preexcited molecules to be

Figure 4. IRMPD yield as a function of effective dissociation fluence for room-temperature SiHCl3 (triangles) and for SiHCl3 preexcited to the 2ν1 level (squares), left-hand scale. The dissociation selectivity, which is the ratio of the two yields, is also plotted as a function of the fluence (open circles, right-hand scale). The NH3 dissociation laser is tuned to 780 cm-1, and the sample pressure is 0.5 Torr.

isotopically selective, since the vibration that we use for IRMPE, ν4, should also have some isotopic shift. We do not expect this selectivity to be significant, however, because the ν4 absorption spectra of the two isotopic species in vibrationally preexcited room temperature molecules are strongly overlapped due to both rotational structure and vibrational statistical (inhomogeneous) broadening.23 We can thus consider the dissociation selectivity plotted in Figure 4 as an estimate for the upper limit of isotopic selectivity in the overall OP-IRMPD process. At an NH3 laser fluence of ∼1.5 J/cm2, where the dissociation yield of the preexcited SiHCl3 is significant, this selectivity is only a factor of 5-6. This can be increased if one preexcites a higher fraction of molecules, but for an optical scheme with unfocused laser beams, which would be best from a practical point of view, the preexcitation fluence is limited to a few J/cm2 because of the damage threshold of IR optical materials. Another way to increase the selectivity is to suppress the dissociation of the ground-state molecules and to favor that of the preexcited species by shifting the dissociation frequency further to the red from the ν4 fundamental. Using the lowest frequency line available from our NH3 laser (773 cm-1) results in an almost 2-fold increase in dissociation yield of the preexcited molecules compared to a 25-30% drop in the dissociation yield of the ground-state species. As a result, this change of dissociation frequency increases the selectivity by a factor of ∼2.5. Figure 5 shows the dependence of the SiCl2 LIF signals on SiHCl3 pressure for both preexcited and room-temperature species dissociated by 1.5 J/cm2 pulses of the ammonia laser tuned to 773 cm-1. Because of collisional quenching of electronically excited SiCl2, the LIF signal increases more slowly with pressure than the dissociation yield. The measured curves thus provide a lower limit for dissociation yield, especially for high pressure. However, since the change of LIF detection efficiency with pressure is the same for SiCl2 produced by dissociation of ground-state and preexcited species, collisional quenching should not influence our measurements of dissociation selectivity, plotted in Figure 5 (right-hand scale). The selectivity reaches a maximum at ∼0.5 Torr and then slowly drops with further increase of pressure, which is likely due to collisional vibrational deactivation of preexcited molecules. The initial increase of the selectivity can be explained by collisional rotational relaxation of a few states initially prepared by

Infrared Laser Chemistry of Trichlorosilane

Figure 5. IRMPD yield as a function of SiHCl3 pressure for the roomtemperature molecules (triangles) and for the species preexcited to the 2ν1 level (squares), left-hand scale. Dissociation selectivity, which is the ratio of the two yields, is also plotted as a function of pressure (open circles, right-hand scale). The NH3 laser is tuned to 773 cm-1, and the dissociation fluence is 1.5 J/cm2.

preexcitation. This thermalization may populate rotational states that are favored for IRMPD by the 773 cm-1 laser line, leading to an increased dissociation yield for the preexcited species. A selectivity of around 10 reached for the sample pressure of 1-2 Torr implies, for example, a maximum possible isotopic enrichment of 29Si species to 35%. If the preexcitation fluence were doubled it will be still below the damage threshold for IR optics, but should increase the level of enrichment to 50%. The dissociation selectivity can be improved by cooling the molecules. Indeed, at room temperature, 82% of SiHCl3 molecules have thermal excitation in low-frequency vibrations (ν3 ) 257.7 cm-1, ν6 ) 175.5 cm-1), giving rise to “hot” bands in the IR absorption spectrum of the ν4. Multiphoton excitation through these “hot” bands may contribute significantly to dissociation of SiHCl3 molecules without preexcitation. Lowering the working temperature will suppress this contribution and substantially increase the dissociation selectivity. Collecting the SiCl2 dissociation fragments in the gas phase requires the use of a scavenger molecule. Addition of a scavenger increases the working pressure and may influence the isotopic selectivity. In this regard, we have performed measurements of isotopic selectivity with He as a buffer gas. Addition of 9 Torr of He to 2 Torr of SiHCl3 already decreases the selectivity by a factor of 5 due to an increased dissociation yield for the nonpreexcited molecules. In general, different collision partners will change this yield differently. Therefore, measurements of the selectivity should be performed for particular scavengers. Below, we report our results of a search for scavengers that could efficiently convert SiCl2 fragments created by IRMPD to stable molecules, although we have not determined the selectivity achievable in the presence of these scavengers. B. IR-Laser-Induced Chemistry of SiHCl3. For IRMPD of room-temperature SiHCl3 we employ an NH3 laser tuned to the 797 cm-1 line with a dissociation fluence at the beam waist of 4-6 J/cm2. Careful measurements show that the absolute IRMPD yield depends linearly on the sample pressure for pressures below 1 Torr. This implies that in this range, dissociation is not influenced by collisional effects (i.e., vibrational deactivation, rotational thermalization). We thus use a working pressure of 0.5 Torr whenever we want to ensure collision-free IRMPD. For sample pressures up to 10 Torr, the

J. Phys. Chem. A, Vol. 107, No. 41, 2003 8581 only gas-phase IRMPD product of pure trichlorosilane that we detect is HCl, in an amount equal to that of dissociated SiHCl3. This is consistent with the lowest dissociation pathway for SiHCl3, as described by eq 1. Over a long irradiation time we observe a solid deposit building up on the reactor windows, and its broad (∼50 cm-1) absorption band centered at 602 cm-1 suggests that this comes from (SiCl2)n polymers. An efficient gas scavenger must therefore suppress the SiCl2 polymerization and convert the radicals to a stable gas-phase product. Moreover, there are two additional requirements that a sucessful scavenger must meet in order to avoid scrambling the isotopic selectivity generated by the laser dissociation processsit should not react directly with the SiHCl3 parent compound and it should not have an IR absorption band that overlaps with the frequency of the dissociation laser. Using the available chemical and spectroscopic data to judge their suitability, we selected a number of candidates for testing as scavengers of SiCl2: (a) di- and tri-atomics: O2, CO2, OCS; (b) halogenated methanes: CH2F2, CF4, CF2Cl2, CH2Br2, CHBr3, CH3I; (c) others: C2H4, C2F4, BCl3, TiCl4, furan (C4H4O). Each of these compounds has been mixed with trichlorosilane in the reaction cell and irradiated by the NH3 laser, and the chemical composition of the reaction products were analyzed spectrophotometrically. The laser-induced reaction of trichlorosilane with the smallest scavenger candidate, O2, exhibits a characteristic threshold dependence on SiHCl3 pressure. Irradiation of 20 Torr of O2 with more than 4 Torr of SiHCl3 results in a visible flash and complete disappearance of trichlorosilane after a single laser shot together with the formation of a film on the windows that exhibits broad IR absorption bands, which we attribute to silicon oxides. We also detect HCl, (SiCl3)2O, SiCl4, and some nonidentified molecule(s) with a siloxane-type Si-O-Si bond. At SiHCl3 pressure below 1 Torr, the bright flash disappears and the dissociation yield drops to ∼0.7% per laser shot. We detect no gaseous products under these conditions. These results are similar to those from thermal and photochemical oxidation of SiHCl3.24 It is likely that at a sufficiently high SiHCl3 pressure, the laser radiation vibrationally heats the molecules to a temperature above the threshold where the chain oxidation reaction occurs. Tests of CO2, OCS, CH2F2, CF4, C2F2, and C2F4 did not reveal detectable gaseous products when 6-90 Torr is mixed with 6 Torr of SiHCl3 and the latter is irradiated with the ammonia laser. Various stable gaseous products have been detected, however, after IRMPD of SiHCl3 mixed separately with CF2Cl2, CH2Br2, CHBr3, CH3I, BCl3, TiCl4, and C4H4O. Table 1 summarizes these products for different mixtures of the tested scavengers with trichlorosilane. Among all the tested compounds, CF2Cl2 seems to be the best suited for scavenging SiCl2 fragments. Irradiation of a mixture of CF2Cl2 with trichlorosilane favors formation of SiCl4 and SiFCl3, although several other compounds are also detected (Table 1). Figure 6 shows the yield of some of the detected products as a function of CF2Cl2 pressure. The yield of SiCl4 is higher than the other products over the entire pressure range and at 60 Torr of CF2Cl2 reaches ∼70% of SiCl3H dissociated by each laser shot in the cell irradiated volume. Thus, at CF2Cl2 pressures in this range, most of the SiCl2 fragments are collected as a single chemical species. Whether or not CF2Cl2 can be used for isotope separation depends on whether it scrambles significantly the isotopic selectivity of the overall process. This is largely determined by the mechanism of the SiCl2 + CF2Cl2 reaction and by collisional

8582 J. Phys. Chem. A, Vol. 107, No. 41, 2003

Figure 6. Yields of different gas products as a function of CF2Cl2 pressure in IRMPD of SiHCl3 mixed with CF2Cl2: 1 (open circles) SiCl4, 2 (closed circles) - SiFCl3, 3 (open squares) - SiF2Cl2, 4 (triangles) - SiF3Cl. The yields are normalized to the dissociation yield of SiHCl3 per each laser shot in the cell irradiated volume.

TABLE 1: IR Photolysis of SiHCl3 Mixed with Different Scavengers SiHCl3 scavenger pressure, pressure, Torr no. scavenger Torr 1

CF2Cl2

0.5 6

2

CH2Br2

0.5 6

3

CH3I

0.5 4

4

BCl3

5

TiCl4

0.5 6 6

6

C4H4O

0.5 2.5

gaseous products

2-20 3-60

none SiCl4, SiFCl3, SiF2Cl2, SiF3Cl, C2F4, HCl 3 none 4-36 presumably, SiBr4-nCln n ) 1-3, C2H4, C2H2, HCl 20 none 20 presumably, SiI4-nCln n ) 2-3, C2H4, CH4, HCl 5-27 SiCl3BCl2 20-212 SiCl4, SiCl3BCl2, HCl 6-15 SiCl4, presumably deposit of TiCl3 and TiCl2 15 C2H2, unidentified IR absorption band at 633 cm-1 18 C2H4, C2H2, unidentified IR absorption bands at 1304, 908, and 633 cm-1

vibrational energy transfer processes. It is known that addition and insertion reactions are the main types of SiCl2 radical reactions with stable compounds.16 However, only with BCl3 do we unambiguously find the insertion reaction product, SiCl3BCl2 (Table 1), identified by its characteristic IR absorption spectrum.25 The mechanism of reaction of SiCl2 with other tested scavengers under laser radiation does not appear to be a simple addition or insertion. We suggest that a common feature of these reactions is the initial thermal decomposition of the scavenger during laser pulse via vibrational energy transfer from highly excited SiHCl3. Indeed, with a typical V-V′ energy transfer rate26 k ∼ 106 s-1 Torr-1, a laser pulse duration τ ∼ 10-6 s and a total pressure p ∼ 10 Torr, the collisional exchange parameter, kτp, is approximately equal to 10. Under these conditions we expect substantial vibrational energy transfer from highly excited SiHCl3 to other molecules. Moreover, a fraction of these species that do not reach dissociation threshold during laser pulse will dissipate their vibrational energy in post-pulse collisions. We

Polianski et al. believe that this transferred vibrational energy is sufficent to dissociate some fraction of the scavenger molecules. The importance of free radicals created by decomposition of a scavenger in binding SiCl2 fragments is supported by the following experiments. First, an increase of a scavenger pressure (e.g., CF2Cl2 or CH2Br2) results in an increase in SiHCl3 dissociation. Normally, the addition of a polyatomic gas results in a drop of dissociation yield due to collisional deactivation of the laser-excited species. In principle, the observed increase of dissociation yield could be explained by collisional refilling of a rotational bottleneck.15 However, substitution of the scavengers with nonreactive Xe and N2 at moderate pressures results in a reduced SiHCl3 dissociation yield, despite the fact that these two gases are highly efficient for rotational relaxation.15 This prompts us to conclude that rotational relaxation is not important at typical pressures of our experiments. Moreover, it leads us to suggest that the observed growth of the dissociation yield with increasing pressure of a scavenger could be due to the reaction of SiHCl3 with fragments created by collision-induced decomposition of the scavenger. A second result in support of this suggestion is the observation that a high yield of the laser-induced reaction is observed only for the scavengers that have a dissociation limit near or lower than that of SiHCl3 (D0 ) 72 kcal/mol18). It is simply difficult to dissociate a scavenger in collisions with molecules that have little vibrational energy. Two more sets of experiments have been performed to clarify the mechanism of the laser-induced reaction. In one, vibrational heating of scavengers in collisions with SiHCl3 has been strongly suppressed by lowering the pressure of trichlorosilane by more than an order of magnitudesdown to 0.5 Torr. No gaseous products were detected under these conditions for any scavenger (see Table 1) except BCl3, where we detect the same insertion productsSiCl3BCl2. In another experiment we have performed IRMPD of CF2Cl2 by a CO2 laser through the lowest dissociation channel of these molecules:15

CF2Cl2 + nhν f CF2Cl + Cl

(2)

These experiments were carried out under conditions where the CF2Cl2 pressure (0.35 and 0.7 Torr) is much smaller than that of SiHCl3 (6 Torr). The SiHCl3 molecules do not absorb radiation from the CO2-laser, and collisional energy transfer from the small number of excited CF2Cl2 to SiHCl3 is negligible. Therefore, trichlorosilane cannot dissociate under these conditions by either direct IRMPD or collisional heating from hot trichlorosilane. Nevertheless, IRMPD of CF2Cl2 in mixtures with SiHCl3 results in an approximately equal consumption of both species, and it is equal to the yield of IRMPD for pure CF2Cl2 (reaction 2). The resulting gaseous products, SiCl4, SiFCl3, and C2F4, are the same as those detected upon IRMPD of 6 Torr of SiHCl3 mixed with an excess CF2Cl2 (as discribed above). Both results support our proposed mechanism of the laserinduced reaction of the investigated scavengers in mixtures with SiHCl3. When more than 3-6 Torr of trichlorosilane in a mixture with a scavenger is irradiated by the ammonia laser, the excited SiHCl3 heats and dissociates the scavenger in collisions. Dissociation fragments of the scavenger may react with SiCl2 radicals or with the SiHCl3 starting material, the latter of which will scramble the isotopic selectivity gained by laser overtone preexcitation of SiHCl3. This leads us to conclude that at moderate trichlorosilane pressure, none of the tested compounds, and in particular CF2Cl2, can be used as an SiCl2 scavenger in an IRMPD-based isotope separation process. At low trichlorosilane pressure (